[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71APKS20FDA5T2D6KA3FEKX":3,"skill-same:01M71APKS20FDA5T2D6KA3FEKX":103},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":65,"commitSha":66,"skillDirectory":67,"forksCount":14,"repositoryHtmlUrl":68,"repositoryHomepageUrl":6,"ownerHtmlUrl":69,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":70,"install":71,"resources":84},"01M71APKS20FDA5T2D6KA3FEKX","fine-tuning-with-trl",null,"Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO\u002FGRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.","sapsapshen\u002Fai-memory-router","sapsapshen","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F20347959?v=4","01M6XDQEF1DP50DKF37H81J3J5","shenmian","01M6XDQEJPE41Y3DT2763X8ZJN",0,"MIT","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Ftrl-fine-tuning\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"tuning","TRL","with","fine tuning","fine tuning with trl","Post-Training","learning","Optimization","for","orchestra","feedback","Transformers","human","llms","Accelerate","when","training","align","model","alignment","2026-11-10 10:12:30",false,"---\nname: fine-tuning-with-trl\ndescription: Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO\u002FGRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.\nversion: 1.0.0\nauthor: Orchestra Research\nlicense: MIT\ndependencies: [trl, transformers, datasets, peft, accelerate, torch]\nmetadata:\n  hermes:\n    tags: [Post-Training, TRL, Reinforcement Learning, Fine-Tuning, SFT, DPO, PPO, GRPO, RLHF, Preference Alignment, HuggingFace]\n\n---\n\n# TRL - Transformer Reinforcement Learning\n\n## Quick start\n\nTRL provides post-training methods for aligning language models with human preferences.\n\n**Installation**:\n```bash\npip install trl transformers datasets peft accelerate\n```\n\n**Supervised Fine-Tuning** (instruction tuning):\n```python\nfrom trl import SFTTrainer\n\ntrainer = SFTTrainer(\n    model=\"Qwen\u002FQwen2.5-0.5B\",\n    train_dataset=dataset,  # Prompt-completion pairs\n)\ntrainer.train()\n```\n\n**DPO** (align with preferences):\n```python\nfrom trl import DPOTrainer, DPOConfig\n\nconfig = DPOConfig(output_dir=\"model-dpo\", beta=0.1)\ntrainer = DPOTrainer(\n    model=model,\n    args=config,\n    train_dataset=preference_dataset,  # chosen\u002Frejected pairs\n    processing_class=tokenizer\n)\ntrainer.train()\n```\n\n## Common workflows\n\n### Workflow 1: Full RLHF pipeline (SFT → Reward Model → PPO)\n\nComplete pipeline from base model to human-aligned model.\n\nCopy this checklist:\n\n```\nRLHF Training:\n- [ ] Step 1: Supervised fine-tuning (SFT)\n- [ ] Step 2: Train reward model\n- [ ] Step 3: PPO reinforcement learning\n- [ ] Step 4: Evaluate aligned model\n```\n\n**Step 1: Supervised fine-tuning**\n\nTrain base model on instruction-following data:\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nfrom trl import SFTTrainer, SFTConfig\nfrom datasets import load_dataset\n\n# Load model\nmodel = AutoModelForCausalLM.from_pretrained(\"Qwen\u002FQwen2.5-0.5B\")\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen\u002FQwen2.5-0.5B\")\n\n# Load instruction dataset\ndataset = load_dataset(\"trl-lib\u002FCapybara\", split=\"train\")\n\n# Configure training\ntraining_args = SFTConfig(\n    output_dir=\"Qwen2.5-0.5B-SFT\",\n    per_device_train_batch_size=4,\n    num_train_epochs=1,\n    learning_rate=2e-5,\n    logging_steps=10,\n    save_strategy=\"epoch\"\n)\n\n# Train\ntrainer = SFTTrainer(\n    model=model,\n    args=training_args,\n    train_dataset=dataset,\n    tokenizer=tokenizer\n)\ntrainer.train()\ntrainer.save_model()\n```\n\n**Step 2: Train reward model**\n\nTrain model to predict human preferences:\n\n```python\nfrom transformers import AutoModelForSequenceClassification\nfrom trl import RewardTrainer, RewardConfig\n\n# Load SFT model as base\nmodel = AutoModelForSequenceClassification.from_pretrained(\n    \"Qwen2.5-0.5B-SFT\",\n    num_labels=1  # Single reward score\n)\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen2.5-0.5B-SFT\")\n\n# Load preference data (chosen\u002Frejected pairs)\ndataset = load_dataset(\"trl-lib\u002Fultrafeedback_binarized\", split=\"train\")\n\n# Configure training\ntraining_args = RewardConfig(\n    output_dir=\"Qwen2.5-0.5B-Reward\",\n    per_device_train_batch_size=2,\n    num_train_epochs=1,\n    learning_rate=1e-5\n)\n\n# Train reward model\ntrainer = RewardTrainer(\n    model=model,\n    args=training_args,\n    processing_class=tokenizer,\n    train_dataset=dataset\n)\ntrainer.train()\ntrainer.save_model()\n```\n\n**Step 3: PPO reinforcement learning**\n\nOptimize policy using reward model:\n\n```bash\npython -m trl.scripts.ppo \\\n    --model_name_or_path Qwen2.5-0.5B-SFT \\\n    --reward_model_path Qwen2.5-0.5B-Reward \\\n    --dataset_name trl-internal-testing\u002Fdescriptiveness-sentiment-trl-style \\\n    --output_dir Qwen2.5-0.5B-PPO \\\n    --learning_rate 3e-6 \\\n    --per_device_train_batch_size 64 \\\n    --total_episodes 10000\n```\n\n**Step 4: Evaluate**\n\n```python\nfrom transformers import pipeline\n\n# Load aligned model\ngenerator = pipeline(\"text-generation\", model=\"Qwen2.5-0.5B-PPO\")\n\n# Test\nprompt = \"Explain quantum computing to a 10-year-old\"\noutput = generator(prompt, max_length=200)[0][\"generated_text\"]\nprint(output)\n```\n\n### Workflow 2: Simple preference alignment with DPO\n\nAlign model with preferences without reward model.\n\nCopy this checklist:\n\n```\nDPO Training:\n- [ ] Step 1: Prepare preference dataset\n- [ ] Step 2: Configure DPO\n- [ ] Step 3: Train with DPOTrainer\n- [ ] Step 4: Evaluate alignment\n```\n\n**Step 1: Prepare preference dataset**\n\nDataset format:\n```json\n{\n  \"prompt\": \"What is the capital of France?\",\n  \"chosen\": \"The capital of France is Paris.\",\n  \"rejected\": \"I don't know.\"\n}\n```\n\nLoad dataset:\n```python\nfrom datasets import load_dataset\n\ndataset = load_dataset(\"trl-lib\u002Fultrafeedback_binarized\", split=\"train\")\n# Or load your own\n# dataset = load_dataset(\"json\", data_files=\"preferences.json\")\n```\n\n**Step 2: Configure DPO**\n\n```python\nfrom trl import DPOConfig\n\nconfig = DPOConfig(\n    output_dir=\"Qwen2.5-0.5B-DPO\",\n    per_device_train_batch_size=4,\n    num_train_epochs=1,\n    learning_rate=5e-7,\n    beta=0.1,  # KL penalty strength\n    max_prompt_length=512,\n    max_length=1024,\n    logging_steps=10\n)\n```\n\n**Step 3: Train with DPOTrainer**\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nfrom trl import DPOTrainer\n\nmodel = AutoModelForCausalLM.from_pretrained(\"Qwen\u002FQwen2.5-0.5B-Instruct\")\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen\u002FQwen2.5-0.5B-Instruct\")\n\ntrainer = DPOTrainer(\n    model=model,\n    args=config,\n    train_dataset=dataset,\n    processing_class=tokenizer\n)\n\ntrainer.train()\ntrainer.save_model()\n```\n\n**CLI alternative**:\n```bash\ntrl dpo \\\n    --model_name_or_path Qwen\u002FQwen2.5-0.5B-Instruct \\\n    --dataset_name argilla\u002FCapybara-Preferences \\\n    --output_dir Qwen2.5-0.5B-DPO \\\n    --per_device_train_batch_size 4 \\\n    --learning_rate 5e-7 \\\n    --beta 0.1\n```\n\n### Workflow 3: Memory-efficient online RL with GRPO\n\nTrain with reinforcement learning using minimal memory.\n\nCopy this checklist:\n\n```\nGRPO Training:\n- [ ] Step 1: Define reward function\n- [ ] Step 2: Configure GRPO\n- [ ] Step 3: Train with GRPOTrainer\n```\n\n**Step 1: Define reward function**\n\n```python\ndef reward_function(completions, **kwargs):\n    \"\"\"\n    Compute rewards for completions.\n\n    Args:\n        completions: List of generated texts\n\n    Returns:\n        List of reward scores (floats)\n    \"\"\"\n    rewards = []\n    for completion in completions:\n        # Example: reward based on length and unique words\n        score = len(completion.split())  # Favor longer responses\n        score += len(set(completion.lower().split()))  # Reward unique words\n        rewards.append(score)\n    return rewards\n```\n\nOr use a reward model:\n```python\nfrom transformers import pipeline\n\nreward_model = pipeline(\"text-classification\", model=\"reward-model-path\")\n\ndef reward_from_model(completions, prompts, **kwargs):\n    # Combine prompt + completion\n    full_texts = [p + c for p, c in zip(prompts, completions)]\n    # Get reward scores\n    results = reward_model(full_texts)\n    return [r[\"score\"] for r in results]\n```\n\n**Step 2: Configure GRPO**\n\n```python\nfrom trl import GRPOConfig\n\nconfig = GRPOConfig(\n    output_dir=\"Qwen2-GRPO\",\n    per_device_train_batch_size=4,\n    num_train_epochs=1,\n    learning_rate=1e-5,\n    num_generations=4,  # Generate 4 completions per prompt\n    max_new_tokens=128\n)\n```\n\n**Step 3: Train with GRPOTrainer**\n\n```python\nfrom datasets import load_dataset\nfrom trl import GRPOTrainer\n\n# Load prompt-only dataset\ndataset = load_dataset(\"trl-lib\u002Ftldr\", split=\"train\")\n\ntrainer = GRPOTrainer(\n    model=\"Qwen\u002FQwen2-0.5B-Instruct\",\n    reward_funcs=reward_function,  # Your reward function\n    args=config,\n    train_dataset=dataset\n)\n\ntrainer.train()\n```\n\n**CLI**:\n```bash\ntrl grpo \\\n    --model_name_or_path Qwen\u002FQwen2-0.5B-Instruct \\\n    --dataset_name trl-lib\u002Ftldr \\\n    --output_dir Qwen2-GRPO \\\n    --num_generations 4\n```\n\n## When to use vs alternatives\n\n**Use TRL when:**\n- Need to align model with human preferences\n- Have preference data (chosen\u002Frejected pairs)\n- Want to use reinforcement learning (PPO, GRPO)\n- Need reward model training\n- Doing RLHF (full pipeline)\n\n**Method selection**:\n- **SFT**: Have prompt-completion pairs, want basic instruction following\n- **DPO**: Have preferences, want simple alignment (no reward model needed)\n- **PPO**: Have reward model, need maximum control over RL\n- **GRPO**: Memory-constrained, want online RL\n- **Reward Model**: Building RLHF pipeline, need to score generations\n\n**Use alternatives instead:**\n- **HuggingFace Trainer**: Basic fine-tuning without RL\n- **Axolotl**: YAML-based training configuration\n- **LitGPT**: Educational, minimal fine-tuning\n- **Unsloth**: Fast LoRA training\n\n## Common issues\n\n**Issue: OOM during DPO training**\n\nReduce batch size and sequence length:\n```python\nconfig = DPOConfig(\n    per_device_train_batch_size=1,  # Reduce from 4\n    max_length=512,  # Reduce from 1024\n    gradient_accumulation_steps=8  # Maintain effective batch\n)\n```\n\nOr use gradient checkpointing:\n```python\nmodel.gradient_checkpointing_enable()\n```\n\n**Issue: Poor alignment quality**\n\nTune beta parameter:\n```python\n# Higher beta = more conservative (stays closer to reference)\nconfig = DPOConfig(beta=0.5)  # Default 0.1\n\n# Lower beta = more aggressive alignment\nconfig = DPOConfig(beta=0.01)\n```\n\n**Issue: Reward model not learning**\n\nCheck loss type and learning rate:\n```python\nconfig = RewardConfig(\n    learning_rate=1e-5,  # Try different LR\n    num_train_epochs=3  # Train longer\n)\n```\n\nEnsure preference dataset has clear winners:\n```python\n# Verify dataset\nprint(dataset[0])\n# Should have clear chosen > rejected\n```\n\n**Issue: PPO training unstable**\n\nAdjust KL coefficient:\n```python\nconfig = PPOConfig(\n    kl_coef=0.1,  # Increase from 0.05\n    cliprange=0.1  # Reduce from 0.2\n)\n```\n\n## Advanced topics\n\n**SFT training guide**: See [references\u002Fsft-training.md](references\u002Fsft-training.md) for dataset formats, chat templates, packing strategies, and multi-GPU training.\n\n**DPO variants**: See [references\u002Fdpo-variants.md](references\u002Fdpo-variants.md) for IPO, cDPO, RPO, and other DPO loss functions with recommended hyperparameters.\n\n**Reward modeling**: See [references\u002Freward-modeling.md](references\u002Freward-modeling.md) for outcome vs process rewards, Bradley-Terry loss, and reward model evaluation.\n\n**Online RL methods**: See [references\u002Fonline-rl.md](references\u002Fonline-rl.md) for PPO, GRPO, RLOO, and OnlineDPO with detailed configurations.\n\n## Hardware requirements\n\n- **GPU**: NVIDIA (CUDA required)\n- **VRAM**: Depends on model and method\n  - SFT 7B: 16GB (with LoRA)\n  - DPO 7B: 24GB (stores reference model)\n  - PPO 7B: 40GB (policy + reward model)\n  - GRPO 7B: 24GB (more memory efficient)\n- **Multi-GPU**: Supported via `accelerate`\n- **Mixed precision**: BF16 recommended (A100\u002FH100)\n\n**Memory optimization**:\n- Use LoRA\u002FQLoRA for all methods\n- Enable gradient checkpointing\n- Use smaller batch sizes with gradient accumulation\n\n## Resources\n\n- Docs: https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Ftrl\u002F\n- GitHub: https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl\n- Papers:\n  - \"Training language models to follow instructions with human feedback\" (InstructGPT, 2022)\n  - \"Direct Preference Optimization: Your Language Model is Secretly a Reward Model\" (DPO, 2023)\n  - \"Group Relative Policy Optimization\" (GRPO, 2024)\n- Examples: https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl\u002Ftree\u002Fmain\u002Fexamples\u002Fscripts\n\n\n\n","# TRL - Transformer Reinforcement Learning\n\n## Quick start\n\nTRL provides post-training methods for aligning language models with human preferences.\n\n**Installation**:\n```bash\npip install trl transformers datasets peft accelerate\n```\n\n**Supervised Fine-Tuning** (instruction tuning):\n```python\nfrom trl import SFTTrainer\n\ntrainer = SFTTrainer(\n    model=\"Qwen\u002FQwen2.5-0.5B\",\n    train_dataset=dataset,  # Prompt-completion pairs\n)\ntrainer.train()\n```\n\n**DPO** (align with preferences):\n```python\nfrom trl import DPOTrainer, DPOConfig\n\nconfig = DPOConfig(output_dir=\"model-dpo\", beta=0.1)\ntrainer = DPOTrainer(\n    model=model,\n    args=config,\n    train_dataset=preference_dataset,  # chosen\u002Frejected pairs\n    processing_class=tokenizer\n)\ntrainer.train()\n```\n\n## Common workflows\n\n### Workflow 1: Full RLHF pipeline (SFT → Reward Model → PPO)\n\nComplete pipeline from base model to human-aligned model.\n\nCopy this checklist:\n\n```\nRLHF Training:\n- [ ] Step 1: Supervised fine-tuning (SFT)\n- [ ] Step 2: Train reward model\n- [ ] Step 3: PPO reinforcement learning\n- [ ] Step 4: Evaluate aligned model\n```\n\n**Step 1: Supervised fine-tuning**\n\nTrain base model on instruction-following data:\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nfrom trl import SFTTrainer, SFTConfig\nfrom datasets import load_dataset\n\n# Load model\nmodel = AutoModelForCausalLM.from_pretrained(\"Qwen\u002FQwen2.5-0.5B\")\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen\u002FQwen2.5-0.5B\")\n\n# Load instruction dataset\ndataset = load_dataset(\"trl-lib\u002FCapybara\", split=\"train\")\n\n# Configure training\ntraining_args = SFTConfig(\n    output_dir=\"Qwen2.5-0.5B-SFT\",\n    per_device_train_batch_size=4,\n    num_train_epochs=1,\n    learning_rate=2e-5,\n    logging_steps=10,\n    save_strategy=\"epoch\"\n)\n\n# Train\ntrainer = SFTTrainer(\n    model=model,\n    args=training_args,\n    train_dataset=dataset,\n    tokenizer=tokenizer\n)\ntrainer.train()\ntrainer.save_model()\n```\n\n**Step 2: Train reward model**\n\nTrain model to predict human preferences:\n\n```python\nfrom transformers import AutoModelForSequenceClassification\nfrom trl import RewardTrainer, RewardConfig\n\n# Load SFT model as base\nmodel = AutoModelForSequenceClassification.from_pretrained(\n    \"Qwen2.5-0.5B-SFT\",\n    num_labels=1  # Single reward score\n)\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen2.5-0.5B-SFT\")\n\n# Load preference data (chosen\u002Frejected pairs)\ndataset = load_dataset(\"trl-lib\u002Fultrafeedback_binarized\", split=\"train\")\n\n# Configure training\ntraining_args = RewardConfig(\n    output_dir=\"Qwen2.5-0.5B-Reward\",\n    per_device_train_batch_size=2,\n    num_train_epochs=1,\n    learning_rate=1e-5\n)\n\n# Train reward model\ntrainer = RewardTrainer(\n    model=model,\n    args=training_args,\n    processing_class=tokenizer,\n    train_dataset=dataset\n)\ntrainer.train()\ntrainer.save_model()\n```\n\n**Step 3: PPO reinforcement learning**\n\nOptimize policy using reward model:\n\n```bash\npython -m trl.scripts.ppo \\\n    --model_name_or_path Qwen2.5-0.5B-SFT \\\n    --reward_model_path Qwen2.5-0.5B-Reward \\\n    --dataset_name trl-internal-testing\u002Fdescriptiveness-sentiment-trl-style \\\n    --output_dir Qwen2.5-0.5B-PPO \\\n    --learning_rate 3e-6 \\\n    --per_device_train_batch_size 64 \\\n    --total_episodes 10000\n```\n\n**Step 4: Evaluate**\n\n```python\nfrom transformers import pipeline\n\n# Load aligned model\ngenerator = pipeline(\"text-generation\", model=\"Qwen2.5-0.5B-PPO\")\n\n# Test\nprompt = \"Explain quantum computing to a 10-year-old\"\noutput = generator(prompt, max_length=200)[0][\"generated_text\"]\nprint(output)\n```\n\n### Workflow 2: Simple preference alignment with DPO\n\nAlign model with preferences without reward model.\n\nCopy this checklist:\n\n```\nDPO Training:\n- [ ] Step 1: Prepare preference dataset\n- [ ] Step 2: Configure DPO\n- [ ] Step 3: Train with DPOTrainer\n- [ ] Step 4: Evaluate alignment\n```\n\n**Step 1: Prepare preference dataset**\n\nDataset format:\n```json\n{\n  \"prompt\": \"What is the capital of France?\",\n  \"chosen\": \"The capital of France is Paris.\",\n  \"rejected\": \"I don't know.\"\n}\n```\n\nLoad dataset:\n```python\nfrom datasets import load_dataset\n\ndataset = load_dataset(\"trl-lib\u002Fultrafeedback_binarized\", split=\"train\")\n# Or load your own\n# dataset = load_dataset(\"json\", data_files=\"preferences.json\")\n```\n\n**Step 2: Configure DPO**\n\n```python\nfrom trl import DPOConfig\n\nconfig = DPOConfig(\n    output_dir=\"Qwen2.5-0.5B-DPO\",\n    per_device_train_batch_size=4,\n    num_train_epochs=1,\n    learning_rate=5e-7,\n    beta=0.1,  # KL penalty strength\n    max_prompt_length=512,\n    max_length=1024,\n    logging_steps=10\n)\n```\n\n**Step 3: Train with DPOTrainer**\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nfrom trl import DPOTrainer\n\nmodel = AutoModelForCausalLM.from_pretrained(\"Qwen\u002FQwen2.5-0.5B-Instruct\")\ntokenizer = AutoTokenizer.from_pretrained(\"Qwen\u002FQwen2.5-0.5B-Instruct\")\n\ntrainer = DPOTrainer(\n    model=model,\n    args=config,\n    train_dataset=dataset,\n    processing_class=tokenizer\n)\n\ntrainer.train()\ntrainer.save_model()\n```\n\n**CLI alternative**:\n```bash\ntrl dpo \\\n    --model_name_or_path Qwen\u002FQwen2.5-0.5B-Instruct \\\n    --dataset_name argilla\u002FCapybara-Preferences \\\n    --output_dir Qwen2.5-0.5B-DPO \\\n    --per_device_train_batch_size 4 \\\n    --learning_rate 5e-7 \\\n    --beta 0.1\n```\n\n### Workflow 3: Memory-efficient online RL with GRPO\n\nTrain with reinforcement learning using minimal memory.\n\nCopy this checklist:\n\n```\nGRPO Training:\n- [ ] Step 1: Define reward function\n- [ ] Step 2: Configure GRPO\n- [ ] Step 3: Train with GRPOTrainer\n```\n\n**Step 1: Define reward function**\n\n```python\ndef reward_function(completions, **kwargs):\n    \"\"\"\n    Compute rewards for completions.\n\n    Args:\n        completions: List of generated texts\n\n    Returns:\n        List of reward scores (floats)\n    \"\"\"\n    rewards = []\n    for completion in completions:\n        # Example: reward based on length and unique words\n        score = len(completion.split())  # Favor longer responses\n        score += len(set(completion.lower().split()))  # Reward unique words\n        rewards.append(score)\n    return rewards\n```\n\nOr use a reward model:\n```python\nfrom transformers import pipeline\n\nreward_model = pipeline(\"text-classification\", model=\"reward-model-path\")\n\ndef reward_from_model(completions, prompts, **kwargs):\n    # Combine prompt + completion\n    full_texts = [p + c for p, c in zip(prompts, completions)]\n    # Get reward scores\n    results = reward_model(full_texts)\n    return [r[\"score\"] for r in results]\n```\n\n**Step 2: Configure GRPO**\n\n```python\nfrom trl import GRPOConfig\n\nconfig = GRPOConfig(\n    output_dir=\"Qwen2-GRPO\",\n    per_device_train_batch_size=4,\n    num_train_epochs=1,\n    learning_rate=1e-5,\n    num_generations=4,  # Generate 4 completions per prompt\n    max_new_tokens=128\n)\n```\n\n**Step 3: Train with GRPOTrainer**\n\n```python\nfrom datasets import load_dataset\nfrom trl import GRPOTrainer\n\n# Load prompt-only dataset\ndataset = load_dataset(\"trl-lib\u002Ftldr\", split=\"train\")\n\ntrainer = GRPOTrainer(\n    model=\"Qwen\u002FQwen2-0.5B-Instruct\",\n    reward_funcs=reward_function,  # Your reward function\n    args=config,\n    train_dataset=dataset\n)\n\ntrainer.train()\n```\n\n**CLI**:\n```bash\ntrl grpo \\\n    --model_name_or_path Qwen\u002FQwen2-0.5B-Instruct \\\n    --dataset_name trl-lib\u002Ftldr \\\n    --output_dir Qwen2-GRPO \\\n    --num_generations 4\n```\n\n## When to use vs alternatives\n\n**Use TRL when:**\n- Need to align model with human preferences\n- Have preference data (chosen\u002Frejected pairs)\n- Want to use reinforcement learning (PPO, GRPO)\n- Need reward model training\n- Doing RLHF (full pipeline)\n\n**Method selection**:\n- **SFT**: Have prompt-completion pairs, want basic instruction following\n- **DPO**: Have preferences, want simple alignment (no reward model needed)\n- **PPO**: Have reward model, need maximum control over RL\n- **GRPO**: Memory-constrained, want online RL\n- **Reward Model**: Building RLHF pipeline, need to score generations\n\n**Use alternatives instead:**\n- **HuggingFace Trainer**: Basic fine-tuning without RL\n- **Axolotl**: YAML-based training configuration\n- **LitGPT**: Educational, minimal fine-tuning\n- **Unsloth**: Fast LoRA training\n\n## Common issues\n\n**Issue: OOM during DPO training**\n\nReduce batch size and sequence length:\n```python\nconfig = DPOConfig(\n    per_device_train_batch_size=1,  # Reduce from 4\n    max_length=512,  # Reduce from 1024\n    gradient_accumulation_steps=8  # Maintain effective batch\n)\n```\n\nOr use gradient checkpointing:\n```python\nmodel.gradient_checkpointing_enable()\n```\n\n**Issue: Poor alignment quality**\n\nTune beta parameter:\n```python\n# Higher beta = more conservative (stays closer to reference)\nconfig = DPOConfig(beta=0.5)  # Default 0.1\n\n# Lower beta = more aggressive alignment\nconfig = DPOConfig(beta=0.01)\n```\n\n**Issue: Reward model not learning**\n\nCheck loss type and learning rate:\n```python\nconfig = RewardConfig(\n    learning_rate=1e-5,  # Try different LR\n    num_train_epochs=3  # Train longer\n)\n```\n\nEnsure preference dataset has clear winners:\n```python\n# Verify dataset\nprint(dataset[0])\n# Should have clear chosen > rejected\n```\n\n**Issue: PPO training unstable**\n\nAdjust KL coefficient:\n```python\nconfig = PPOConfig(\n    kl_coef=0.1,  # Increase from 0.05\n    cliprange=0.1  # Reduce from 0.2\n)\n```\n\n## Advanced topics\n\n**SFT training guide**: See [references\u002Fsft-training.md](references\u002Fsft-training.md) for dataset formats, chat templates, packing strategies, and multi-GPU training.\n\n**DPO variants**: See [references\u002Fdpo-variants.md](references\u002Fdpo-variants.md) for IPO, cDPO, RPO, and other DPO loss functions with recommended hyperparameters.\n\n**Reward modeling**: See [references\u002Freward-modeling.md](references\u002Freward-modeling.md) for outcome vs process rewards, Bradley-Terry loss, and reward model evaluation.\n\n**Online RL methods**: See [references\u002Fonline-rl.md](references\u002Fonline-rl.md) for PPO, GRPO, RLOO, and OnlineDPO with detailed configurations.\n\n## Hardware requirements\n\n- **GPU**: NVIDIA (CUDA required)\n- **VRAM**: Depends on model and method\n  - SFT 7B: 16GB (with LoRA)\n  - DPO 7B: 24GB (stores reference model)\n  - PPO 7B: 40GB (policy + reward model)\n  - GRPO 7B: 24GB (more memory efficient)\n- **Multi-GPU**: Supported via `accelerate`\n- **Mixed precision**: BF16 recommended (A100\u002FH100)\n\n**Memory optimization**:\n- Use LoRA\u002FQLoRA for all methods\n- Enable gradient checkpointing\n- Use smaller batch sizes with gradient accumulation\n\n## Resources\n\n- Docs: https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Ftrl\u002F\n- GitHub: https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl\n- Papers:\n  - \"Training language models to follow instructions with human feedback\" (InstructGPT, 2022)\n  - \"Direct Preference Optimization: Your Language Model is Secretly a Reward Model\" (DPO, 2023)\n  - \"Group Relative Policy Optimization\" (GRPO, 2024)\n- Examples: https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl\u002Ftree\u002Fmain\u002Fexamples\u002Fscripts\n\n\n\n",{"name":5,"author":44,"license":15,"version":45,"metadata":46,"description":7,"dependencies":58},"Orchestra Research","1.0.0",{"hermes":47},{"tags":48},[24,20,49,50,51,52,53,54,55,56,57],"Reinforcement Learning","Fine-Tuning","SFT","DPO","PPO","GRPO","RLHF","Preference Alignment","HuggingFace",[59,60,61,62,63,64],"trl","transformers","datasets","peft","accelerate","torch","main","7a38d8e7073e61c16eda85681cb32fb68a8657a5","skills-store\u002Fhermes\u002Ftrl-fine-tuning","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router","https:\u002F\u002Fgithub.com\u002Fsapsapshen","2026-04-14 21:12:12",{"source":8,"skillName":5,"commands":72,"prompt":82,"zipAvailable":83},[73,76,79],{"tool":74,"command":75},"npx","npx skills add sapsapshen\u002Fai-memory-router --skill fine-tuning-with-trl",{"tool":77,"command":78},"bunx","bunx skills add sapsapshen\u002Fai-memory-router --skill fine-tuning-with-trl",{"tool":80,"command":81},"pnpm","pnpm dlx skills add sapsapshen\u002Fai-memory-router --skill fine-tuning-with-trl","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Ftrl-fine-tuning\u002FSKILL.md",true,[85,91,95,99],{"path":86,"resourceType":87,"resourceTypeDesc":88,"sizeBytes":89,"sourceHtmlUrl":90},"references\u002Fdpo-variants.md","REFERENCE","参考文档",4297,"https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Ftrl-fine-tuning\u002Freferences\u002Fdpo-variants.md",{"path":92,"resourceType":87,"resourceTypeDesc":88,"sizeBytes":93,"sourceHtmlUrl":94},"references\u002Fonline-rl.md",1971,"https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Ftrl-fine-tuning\u002Freferences\u002Fonline-rl.md",{"path":96,"resourceType":87,"resourceTypeDesc":88,"sizeBytes":97,"sourceHtmlUrl":98},"references\u002Freward-modeling.md",2597,"https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Ftrl-fine-tuning\u002Freferences\u002Freward-modeling.md",{"path":100,"resourceType":87,"resourceTypeDesc":88,"sizeBytes":101,"sourceHtmlUrl":102},"references\u002Fsft-training.md",3237,"https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Ftrl-fine-tuning\u002Freferences\u002Fsft-training.md",[104,127,152,162,175,194],{"skillNo":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":115,"publishedAt":39},"01M71ATQRKG48SH06FQ697T5G9","wechat-search","Search WeChat Official Account articles using OpenClaw's web search and fetch capabilities with compliance-focused design.","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Fwechat-search-release\u002FSKILL.md",[110],{"categoryNo":111,"categoryName":112,"taxonomyType":113,"taxonomyTypeDesc":114},"01JBSEEDCAT000000000000037","律师与法务","PROFESSION","适用职业",[116,117,118,119,120,121,122,123,124,125,126,21],"wechat search","wechat","search","capabilities","openclaw","design","web search","compliance","web","official","fetch",{"skillNo":128,"name":129,"nameZh":6,"description":130,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":131,"categories":132,"keywords":133,"publishedAt":39},"01M71ATNAVGZB35PFYM2GC5N00","trend-tap","Real-time trending topics aggregator across 7 platforms (X\u002FTwitter, Reddit, Google Trends, Hacker News, Zhihu, Bilibili, Weibo). Trigger: when user says 'trend', 'trends', 'trend tap', '热点', '热搜', or asks what is trending\u002Fhot. Do NOT use web_search — this skill already fetches live data from all platforms with zero API keys.","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Ftrend-tap\u002FSKILL.md",[],[134,135,136,137,138,139,118,140,141,142,143,34,144,145,146,147,148,149,150,151],"trend","trend tap","tap","api keys","Bash","time","trending","hacker","aggregator","do","brew","live","data","google","install","read","across","api",{"skillNo":153,"name":154,"nameZh":6,"description":155,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":156,"categories":157,"keywords":158,"publishedAt":39},"01M71ATJ9AM5KM15AA8PS302EB","trend-radar","Real-time trending topics aggregator across 7 platforms (X\u002FTwitter, Reddit, Google Trends, Hacker News, Zhihu, Bilibili, Weibo). Trigger: when user says 'trend', 'trends', 'trend pulse', '热点', '热搜', or asks what is trending\u002Fhot. Do NOT use web_search — this skill already fetches live data from all platforms with zero API keys.","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Ftrend-radar\u002FSKILL.md",[],[159,160,134,144,145,146,147,138,118,149,161,150,151,34,142,141,140,148,139,137],"trend radar","radar","pulse",{"skillNo":163,"name":164,"nameZh":6,"description":165,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":166,"categories":167,"keywords":168,"publishedAt":39},"01M71ATGF095N3EJM9G8SD0BB5","tech-news-daily","每日科技新闻汇——从全球多个 RSS 源抓取最新科技新闻，翻译（英→中）、去重、排版成每日简报。当用户询问科技新闻、科技新鲜事、每日科技、技术动态、Daily Tech News 或类似表达时触发此技能。","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Ftech-news-daily\u002FSKILL.md",[],[169,170,171,172,173,174],"news","tech","tech news daily","daily","rss","翻译",{"skillNo":176,"name":177,"nameZh":6,"description":178,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":179,"categories":180,"keywords":186,"publishedAt":39},"01M71ATFAEAMQ17M70NQNY7SR8","summarize","Summarize URLs or files with the summarize CLI (web, PDFs, images, audio, YouTube).","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Fsummarize\u002FSKILL.md",[181],{"categoryNo":182,"categoryName":183,"taxonomyType":184,"taxonomyTypeDesc":185},"01JBSEEDCAT000000000000021","资料检索与研究","SCENARIO","使用场景",[177,124,187,188,189,21,144,190,191,192,136,193,148],"Audio","cli","youtube","images","https","files","pdfs",{"skillNo":195,"name":196,"nameZh":6,"description":197,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":198,"categories":199,"keywords":200,"publishedAt":39},"01M71ATDNS3R0M4H936PACQAWZ","sonoscli","Control Sonos speakers (discover\u002Fstatus\u002Fplay\u002Fvolume\u002Fgroup).","https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fopenclaw\u002Fsonoscli\u002FSKILL.md",[],[196,201,202,203,148,204,191,205,206,207,208,209],"status","play","go","latest","github","discover","control","cmd","volume"]